Current triage systems show limited ability to identify critically ill patients. We developed and validated machine-learning methods to predict mortality and hospitalization. A single center, retrospective study was conducted with 164,335 ER visits during 2016~2017. Decision tree and random forest model showed prediction accuracy (Area Under Curve) which is significantly higher than current triage system; Korean Triage and Acuity Scale.
Learning Objective 1: Learn to develop and validate machine-learning methods for triage system in emergency room.
Jaeyong Yu (Presenter)
gabyong jeong, samsung medical center
wonchul cha, samsung medical center